The authors of the target article provide an excellent discussion on the evolution of human intelligence, particularly on the formation of secondary modules that are more variable and domain-general. As discussed in section 1 of the target article, general intelligence seems evolutionarily implausible because the mind is populated by a large number of adaptive specializations that are functionally organized to solve evolutionarily typical and recurrent problems of survival and reproduction (see also Cosmides et al. Reference Cosmides, Barrett and Tooby2010; Wang Reference Wang1996). To resolve this paradox, the authors propose a model that construes the mind as a mix of truly modular skills (primary modules) and more variable and flexible skills (secondary modules) that are ontogenetically acquired via the guidance of general intelligence (see sect. 1.2.3). In the following, I propose a novel hypothesis to extend this discussion by showing that secondary or higher-order modules can be formed not only ontogenetically, but also phylogenetically as adaptations, evolved from domain-specific modules.
If general intelligence consists of a set of secondary modules, each secondary module may be an evolved programing solution for a function that could be shared by multiple primary modules. These secondary modules of general intelligence can be either ontogenetically constructed or phylogenetically evolved. Imagine that a computer architect was creating a system called Unix using the programming language C. At the beginning, the operating system was written in assembly, where nearly every line would contain memory addresses. Would it be possible to program the system for its input/output devices without repetitively stating these tedious memory addresses? This problem has been solved by creating a pointer variable, whose value specifies the address of a memory location. If a memory address is called upon repeatedly, creating a pointer to store the address would be an effective programing solution. Similarly, if a random number generator is used repeatedly by many local modules, it would be more efficient to make it globally accessible by each of the modules.
Now imagine you are using a computer and have created many folders for different papers. At the beginning, you included a copy of a word processor in each folder. You then realized that all of these papers require a word processor. It would be more efficient if you place a single copy of a generic word processor in a visible place that is accessible by all of the papers. This word processor has then become a general tool for a common requirement of different tasks. Similarly, numeracy, as a component of general intelligence, may be evolved as a result of a common demand by multiple specific adaptations (e.g., counting foraging outcomes; gauging social exchanges, assessing mate values, tracking reciprocal activities, etc.). A general-purpose device would be cognitively economical if it is utilized for multiple tasks. From a design viewpoint, general intelligence comes as a solution for overlapping components of primary modules or for coordinating secondary modules via executive functions (see sect. 1.2.2). From this perspective, task-specificity to solve a unique adaptive problem (e.g., foraging, hunting, or mating) should be distinguished from function-specificity to deal with a common computational demand (e.g., numeracy, verbal communication, etc.)
By the same token, if a particular emotion is a common component of many specific adaptations, this basic emotion would become a general mechanism shared by these adaptations. For instance, anger is the expression of a neurocomputational system that evolved to adaptively regulate behavior in the context of resolving conflicts of interest in favor of the angry individual (Cosmides & Tooby Reference Cosmides and Tooby2013). Anger can be triggered by multiple task-specific adaptations, such as territory defense, mating competition, sibling rivalry, and cheater detection. Once triggered, the anger system would produce one of two outputs: threatening to inflict costs (aggression) or threatening to withdraw expected benefits (Cosmides & Tooby Reference Cosmides and Tooby2013). Similarly, fear is a basic emotion that plays a role in multiple adaptations and has its brain center mainly located in the amygdala. This localized brain function allows the organism to react not only to specific and typical fear-inducing stimuli, but also to learn to react to nonspecific stimuli with fear via fear conditioning (e.g., Phelps & LeDoux Reference Phelps and LeDoux2005).
General intelligence and basic emotions may both be solutions for multiple primary modules that demand some common functions. This pointer's hypothesis of general intelligence challenges a couple assumptions in the research literature of cognitive evolution. As indicated by the authors, many previous accounts of evolution of human intelligence assume that domain-specific modules ought to be cheaper and simpler than domain-general cognitive mechanisms (see sect. 1 for relevant discussion). However, being specific does not necessarily mean that the mechanism is simple or cognitively economical. Because a domain-specific mechanism is designed for solving a specific problem, its design purpose is to do whatever it takes to solve the problem instead of achieving structural simplicity, computational economy, or functional efficiency. Such designs can be either as exquisite as the human visual system or as patchy and lousy as a male's reproductive system, revised and modified from the Wolffian duct. Thus, these adaptive specializations can either be cheap and simple or costly and complex. Unlike engineering designs, evolutionary designs cannot afford to erase existing blueprints and start from scratch. Evolutionary efficiency is inevitably an efficiency under phylogenetic constraints.
The pointer's hypothesis also challenges the notion that domain-specific mechanisms are more localized in the brain than domain-general mechanisms (see also sect. 1.1 for relevant discussion). However, this notion is at odds with the following two observations. First, a specific adaptation can be implemented by a distributed neural network. Second, a localized brain function is likely a result of a common demand of multiple primary modules. Thus, a more general-purpose mechanism may be implemented by allocating a particular brain region to perform a function shared by multiple primary modules. For instance, a localized motor cortex (e.g., the precentral gyrus) can be used for motor controls in foraging, hunting, gathering, mating competition, and so on. For the same reason, localized brain regions for language processing serve as a general-purpose system for all of the tasks that require information exchange and verbal communication.
You have
Access
The authors of the target article provide an excellent discussion on the evolution of human intelligence, particularly on the formation of secondary modules that are more variable and domain-general. As discussed in section 1 of the target article, general intelligence seems evolutionarily implausible because the mind is populated by a large number of adaptive specializations that are functionally organized to solve evolutionarily typical and recurrent problems of survival and reproduction (see also Cosmides et al. Reference Cosmides, Barrett and Tooby2010; Wang Reference Wang1996). To resolve this paradox, the authors propose a model that construes the mind as a mix of truly modular skills (primary modules) and more variable and flexible skills (secondary modules) that are ontogenetically acquired via the guidance of general intelligence (see sect. 1.2.3). In the following, I propose a novel hypothesis to extend this discussion by showing that secondary or higher-order modules can be formed not only ontogenetically, but also phylogenetically as adaptations, evolved from domain-specific modules.
If general intelligence consists of a set of secondary modules, each secondary module may be an evolved programing solution for a function that could be shared by multiple primary modules. These secondary modules of general intelligence can be either ontogenetically constructed or phylogenetically evolved. Imagine that a computer architect was creating a system called Unix using the programming language C. At the beginning, the operating system was written in assembly, where nearly every line would contain memory addresses. Would it be possible to program the system for its input/output devices without repetitively stating these tedious memory addresses? This problem has been solved by creating a pointer variable, whose value specifies the address of a memory location. If a memory address is called upon repeatedly, creating a pointer to store the address would be an effective programing solution. Similarly, if a random number generator is used repeatedly by many local modules, it would be more efficient to make it globally accessible by each of the modules.
Now imagine you are using a computer and have created many folders for different papers. At the beginning, you included a copy of a word processor in each folder. You then realized that all of these papers require a word processor. It would be more efficient if you place a single copy of a generic word processor in a visible place that is accessible by all of the papers. This word processor has then become a general tool for a common requirement of different tasks. Similarly, numeracy, as a component of general intelligence, may be evolved as a result of a common demand by multiple specific adaptations (e.g., counting foraging outcomes; gauging social exchanges, assessing mate values, tracking reciprocal activities, etc.). A general-purpose device would be cognitively economical if it is utilized for multiple tasks. From a design viewpoint, general intelligence comes as a solution for overlapping components of primary modules or for coordinating secondary modules via executive functions (see sect. 1.2.2). From this perspective, task-specificity to solve a unique adaptive problem (e.g., foraging, hunting, or mating) should be distinguished from function-specificity to deal with a common computational demand (e.g., numeracy, verbal communication, etc.)
By the same token, if a particular emotion is a common component of many specific adaptations, this basic emotion would become a general mechanism shared by these adaptations. For instance, anger is the expression of a neurocomputational system that evolved to adaptively regulate behavior in the context of resolving conflicts of interest in favor of the angry individual (Cosmides & Tooby Reference Cosmides and Tooby2013). Anger can be triggered by multiple task-specific adaptations, such as territory defense, mating competition, sibling rivalry, and cheater detection. Once triggered, the anger system would produce one of two outputs: threatening to inflict costs (aggression) or threatening to withdraw expected benefits (Cosmides & Tooby Reference Cosmides and Tooby2013). Similarly, fear is a basic emotion that plays a role in multiple adaptations and has its brain center mainly located in the amygdala. This localized brain function allows the organism to react not only to specific and typical fear-inducing stimuli, but also to learn to react to nonspecific stimuli with fear via fear conditioning (e.g., Phelps & LeDoux Reference Phelps and LeDoux2005).
General intelligence and basic emotions may both be solutions for multiple primary modules that demand some common functions. This pointer's hypothesis of general intelligence challenges a couple assumptions in the research literature of cognitive evolution. As indicated by the authors, many previous accounts of evolution of human intelligence assume that domain-specific modules ought to be cheaper and simpler than domain-general cognitive mechanisms (see sect. 1 for relevant discussion). However, being specific does not necessarily mean that the mechanism is simple or cognitively economical. Because a domain-specific mechanism is designed for solving a specific problem, its design purpose is to do whatever it takes to solve the problem instead of achieving structural simplicity, computational economy, or functional efficiency. Such designs can be either as exquisite as the human visual system or as patchy and lousy as a male's reproductive system, revised and modified from the Wolffian duct. Thus, these adaptive specializations can either be cheap and simple or costly and complex. Unlike engineering designs, evolutionary designs cannot afford to erase existing blueprints and start from scratch. Evolutionary efficiency is inevitably an efficiency under phylogenetic constraints.
The pointer's hypothesis also challenges the notion that domain-specific mechanisms are more localized in the brain than domain-general mechanisms (see also sect. 1.1 for relevant discussion). However, this notion is at odds with the following two observations. First, a specific adaptation can be implemented by a distributed neural network. Second, a localized brain function is likely a result of a common demand of multiple primary modules. Thus, a more general-purpose mechanism may be implemented by allocating a particular brain region to perform a function shared by multiple primary modules. For instance, a localized motor cortex (e.g., the precentral gyrus) can be used for motor controls in foraging, hunting, gathering, mating competition, and so on. For the same reason, localized brain regions for language processing serve as a general-purpose system for all of the tasks that require information exchange and verbal communication.